Real-time AI-based quality assurance for semiconductor production machines

2. The method as described in claim 1 wherein the prediction is one of: a probability of a defect on the given product, a predicted value of an end-of-line quality test, a predicted value of an in-line quality test, a predicted value of an integration test, and a predicted value of a field performance with respect to the given product.

3. The method as described in claim 1 wherein the production system is a semiconductor manufacturing execution system (MES).

4. The method as described in claim 1 wherein the quality test data includes labeled and unlabeled data.

5. The method as described in claim 1 wherein the prediction is associated with one of: a plasma etching machine, a chemical vapor deposition (CVD) machine, a chemical mechanical planarization (CMP) machine, and an ion implantation machine.

6. The method as described in claim 1 wherein the transfer learning transforms data associated with a first machine in the production system such that the data appears associated with a second machine in the production system.

7. The method as described in claim 1 wherein the production data comprises non-homogeneous machine parameters and maintenance data from two or more machines of a same physical type.

8. The method as described in claim 1 wherein the production data comprises non-homogeneous machine parameters and maintenance data collected from a single machine over a time period.

9. The method as described in claim 1 wherein the production data is received periodically.

10. The method as described in claim 1 wherein the training and the real-time prediction occur remotely to the production system.

11. The method as described in claim 1 wherein the training occurs remotely to the production system and the real-time prediction is generated and provided on-premises to the production system.

12. The method as described in claim 1 wherein the training and the real-time prediction occur on-premises to the production system.

14. The method as described in claim 13 wherein the prediction is one of: a probability of a defect on the given product, a predicted value of an end-of-line quality test to be performed on the given product, a predicted value of an in-line quality test to be performed on the given production machine, a predicted value of an integration test, and a predicted value of a field performance with respect to the given product.

15. The method as described in claim 13 wherein the neural network is an ensemble of models comprising a neural network that is a common representation of a type of the given production machine, and a neural network that has been trained using production data from another production machine of the type.

16. The method as described in claim 13 wherein the prediction indicates how well an in-line quality test correlates with an end-of-line quality test.

17. The method as described in claim 13 further including periodically re-training the neural network.

18. The method as described in claim 13 further including providing the prediction to facilitate a quality assurance task in the production system.